Psychiatric disorders, as well as normal aging, are often associated with a wide range of cognitive changes. Recent developments in neuroscience, particularly multimodal neuroimaging techniques, can provide better understanding of neural mechanisms that underlie these changes. Mediation analysis is a popularly used statistical method to investigate neural mechanisms. However, existing statistical methods were not designed to accommodate such large-scale, multi-dimensional, and complicated data in mediation analysis. The overarching aim of my K01 Mentored Research Development Award is to acquire training that will allow me to pursue a line of mental health research related to cognition and to develop novel statistical methods to support the emerging research in cognitive neuroimaging and mental health. I propose to receive training in: 1) cognition and its relationship with aging and psychopathology; 2) neural basis in cognition and multimodal neuroimaging; 3) functional mediation and selective mediation analysis. This will provide me with the tools to conduct research that will fill significant gaps in the statistical analyses on cognitive neuroimaging, particularly focusing on aging and neurodegeneration-associated cognitive decline: First, I propose to review and develop analytic tools to examine how aging and psychopathology affect cognitive system constructs. Understanding the patterns of age-related differences in each of these cognitive constructs and the possible heterogeneity related to the psychopathology is of great importance. I will test nonlinearity and heterogeneity in the latent cognitive constructs using two independent large data. Second, I will develop and examine activation networks (reference ability neural networks (RANNs)) associated with each of the cognitive constructs. Although it is critical to understand underlying neural networks of cognition, there are many unsolved problems in investigating the presence or structure of RANNs. I will develop a method to derive underlying networks, find RANNs taking into account possible spatial overlap among RANNs, and investigate the relationship between RANNs expression and performance on the concomitant cognitive constructs. Third, I will develop selective mediation analyses to examine the relationship between imaged neural substrates of aging and psychiatric disorder, such as regional volume loss, and cognitive decline and test if the different imaged brain modalities selectively mediate cognitive aging by influencing the expression of cognitive ability constructs. In totality, this training and research will inform hypotheses for an R01 grant application to be submitted in Year 4 of the award period. While to date I have exemplified a productive research career in biostatistics and psychiatry, I need substantially more training for successful completion of the research proposed in this application and to make a lasting contribution to the field of mental health.